Voice cloning is the synthesis of a voice from a sample. Modern models (ElevenLabs, Resemble, OpenAI) need 30 seconds to 10 minutes of clean audio to produce a synthetic profile that preserves the original voice’s characteristics.
For enterprises this enables a consistent brand voice across every call — independent of the TTS vendor and without falling back on stock voices. Multi-language voice profiles built from a single recording are now possible too.
Legally, voice cloning is sensitive: a natural person’s voice is biometric personal data (GDPR Art. 9). Documented consent, purpose limitation, storage caps and a revocation process are mandatory. Without that framework, deployment is high-risk.